One Research Question, Three Search Strategies: Which Finds More Relevant Papers?

I tested three routes for the same question about digital sleep interventions for university students: a keyword search, a question-shaped search, and citation chasing from a relevant seed paper.

The useful finding was not that one route won. Each route solved a different part of the search.

The keyword query was: “digital cognitive behavioral therapy insomnia university students.” It quickly found a directly related paper and several digital CBT-I studies.

It also drifted. Some papers involved depression, diabetes, or heavy drinking; others were reviews rather than original student trials. The query was precise about the intervention but still ambiguous about population and document type.

Keyword search worked best for speed and reproducibility. Its weakness was vocabulary: a relevant paper that says “internet-delivered,” “email-delivered,” or simply “online sleep program” may not rank as highly.

Instead of a keyword bundle, write the desired evidence in a sentence:

Question-shaped or semantic search can match meaning across different wording. It is especially useful when you know the design and population but not the field's terminology.

The trade-off is transparency. A natural-language request is easy to improve, but harder to reproduce exactly than a documented database search. Save the complete prompt, tool, filters, and date.

I used a seed paper closely matched to the question and checked its references and newer citing work. This route surfaced the underlying randomized trial, related digital CBT research, and a later commentary.

Citation chasing was the cleanest route to the research conversation around that paper. It was also the narrowest. A seed anchors you to its authors' vocabulary, date, and citation neighborhood. One seed is a path, not a map of the field.

Use keywords when you can name the core concepts. Use a full-sentence search when your problem is clearer than your vocabulary. Use citation chasing when you already trust one paper and want its intellectual neighborhood.

For a serious literature review, run them as a loop:

The best search is usually the combination that finds known papers and still surprises you with relevant work from outside your first wording.

Start with the question in WisPaper Search. Save good candidates to My Library, compare their terminology and references, then rerun the search with the population, intervention, or outcome terms you learned.

If the first results are noisy, do not ask for “more.” Change one part of the question and observe which papers enter or leave the set.